Model with conditional interaction
In this example, we show regional effects of a model with conditional interactions using PDP, ALE, and RHALE. In particular, we:
- show how to use
effector
to estimate the regional effects using PDP, ALE, and RHALE - provide the analytical formulas for the regional effects
- test that (1) and (2) match
We will use the following model:
where the features \(x_1, x_2, x_3\) are independent and uniformly distributed in the interval \([-1, 1]\).
The model has an interaction between \(x_1\) and \(x_2\) caused by the terms: \(f_{1,2}(x_1, x_2) = -x_1^2 \mathbb{1}_{x_2 <0} + x_1^2 \mathbb{1}_{x_2 \geq 0}\). This means that the effect of \(x_1\) on the output \(y\) depends on the value of \(x_2\) and vice versa. Therefore, there is no golden standard on how to split the effect of \(f_{1,2}\) to two parts, one that corresponds to \(x_1\) and one to \(x_2\). Each global effect method has a different strategy to handle this issue. Below we will see how PDP, ALE, and RHALE handle this interaction.
In contrast, \(x_3\) does not interact with any other feature, so its effect can be easily computed as \(e^{x_3}\).